Document parsing tags
Tags list OCR, PDF, document parsing and understanding, layout analysis, table and chart parsing, and markdown output.
Open Source Model Profile · infly
Infinity-Parser2-Flash
Infinity-Parser2-Flash is a 2.21B-parameter Qwen document parser from infly. Its card documents OCR, layout, and markdown parsing with low-latency inference.
- Publisher
- infly
- Task
- image-text-to-text
- Model type
- qwen3_5
- License
- apache-2.0
- Library
- transformers
- Publication status
- Accepted · not indexed
Model overview
Infinity-Parser2-Flash is published by infly as an image-text-to-text document-understanding model. The captured configuration identifies Qwen3_5ForConditionalGeneration with 2,213,241,664 parameters under Apache-2.0. According to the model card, it is the low-latency variant of the Infinity-Parser2 family.
Recorded capabilities
Publisher-reported Flash efficiency
According to the model card, the Flash variant is engineered for low-latency inference with a publisher-reported 3.68x speedup over Infinity-Parser-7B.
5M-sample data engine
The model card describes nearly 5 million parsing samples, and captured tags reference dataset infly/Infinity-Doc2-5M.
Parser API and backends
The card documents an InfinityParser2 parsing API with doc2json, doc2md, and custom prompts across vLLM, Transformers, and server backends.
Use cases in the source record
- Document-to-markdown conversion for PDFs and images, including layout, tables, charts, and formula content.
- Batch parsing workflows using the documented InfinityParser2 API with vLLM, Transformers, or hosted-server backends.
Limitations and unknowns
- Benchmark claims such as SOTA on olmOCR-bench and ParseBench and reported ParseBench scores are publisher-reported and have not been independently verified by Ethen.
- No context-window value was extracted from this record; the card's 65,536 and 300,000 token figures describe server and evaluation settings.
- Provider state is historical snapshot data, not independently refreshed current availability.
Source and provenance
Source: infly/Infinity-Parser2-Flash
Captured: Unknown. Processed: 2026-09-07T19:35:57.215137+00:00.
Infinity-Parser2-Flash 💻 Github | 📊 Dataset | 📄 Paper | 🚀 Demo News [2026-05-11] Released flagship document parsing models: Infinity-Parser2-Pro , Infinity-Parser2-Flash , and the dataset Infinity-Doc2-5M . Infinity-Parser2 achieves SOTA results on olmOCR-bench and ParseBench. Introduction We are excited to release Infinity-Parser2, our latest flagship document understanding model. We offer two distinct variants to address diverse deployment constraints: Infinity-Parser2-Pro, optimized for maximum accuracy in precision-critical tasks, achieves state-of-the-art results on olmOCR-Bench (87.6%) and ParseBench (74.3%), surpassing fr…
F001F002F003F004F005F006F007F008F009F010F012F013F014F015F019F020F021F024